Seminars 2025

AIC Lab seminar

Date: 13.03.2025

Speaker: Pavel Sulimov, Senior Quantum AI Researcher at Institute of Computer Science, Zurich University of Applied Sciences

Title: "Quantum Machine Learning in 2025: Myths & Facts"

Abstract: Classical deep learning is currently on a big hype, making a great path in last decade from GANs and AlphaGo to AI models with reasoning. But can we do even better, e.g. by enhancing AI with 100-year-old technology like quantum mechanics? What if we add phenomena like superposition, entanglement, and quantum parallelism into neural network, call it quantum neural network, and ... why should it be better than the classical one?

At the seminar we'll cover the basics of quantum machine learning, understanding the main analogies and differences with classical AI, and look into the advantages (like applications in drug discovery, financial markets, large images analysis etc.) and difficulties (like error mitigation, hardware limitations etc.). Some of the information on the Internet is true, some is not - but don't worry, except from just discovering myths and legends, we'll also "touch" the real IBM quantum computer, and see how it can be used already now.

Video of the seminar by the link. 

Date: 09.04.2025

Speaker: Attila Kertesz-Farkas, Laboratory Head

Title: "False discovery rate control in large-scale hypothesis testing".

Abstract: In the talk we will discuss False Discovery rate (FDR) control procedures, including the Benjamini-Hochberg protocol, the target-decoy competition-based methods, and some more advanced ones for multiple hypothesis testing, when hypothesis can be grouped. We also discuss their statistical power.

Video of the seminar by the Link

Date: 16.04.2025

Speaker: Kishankumar Bhimani, Researcher

Title: "Exact p-value calculation for high resolution Tandem Mass Spectrometry data"

Abstract: Kishan's research talk will highlights, how we identify peptides from high-resolution mass spectrometry data. First, a faster version of the XPV algorithm was developed, and speeding up run-time for peptide identification. Second, the XPV method was adapted to high-resolution data (HR-XPV), improving the accuracy of p-values and reducing false matches. Third, a new algorithm called SeVa was created to detect single amino acid variations, helping to find mutated peptides in complex samples. These advancements make peptide analysis more reliable, especially in large-scale proteomics and MS/MS studies.